Ocean region

Sara Rocha1-2, Miguel A. Carretero', Miguel Vences3, Frank Glaw4 and

D. James Harris1,2*

'Centro de Investigação em Biodiversidade e Recursos Genéticos (CIBIO/UP), ICETA, Campus Agrário de Vairão, Vila do Conde, Portugal, 2Departamento de Zoologia-

Antropologia, Faculdade de Ciências da Universidade do Porto, Porto, Portugal, ^Institute for Biodiversity and Ecosystem Dynamics, Zoological Museum, University of Amsterdam, Amsterdam, The Netherlands and

4Zoologische Staatssammlung, Miinchen,

Germany

'Correspondence: D. James Harris, Centro de Investigação em Biodiversidade e Recursos Genéticos (CIBIO/UP), ICETA, Campus Agrário de Vairão, 4485-661 Vila do Conde, Portugal.

ABSTRACT

Aim Cryptoblepharus is a genus of small arboreal or rock-dwelling scincid lizards, widespread through the Indo-Pacific and Australian regions, with a disjunct outlier in the Malagasy region. The taxonomy within this genus is controversial, with different authors ranking the different forms (now some 36) at various levels, from different species to subspecies of a single species, Cryptoblepharus

boutonii. We investigated the biogeography and genetic differentiation of the Cryptoblepharus from the Western Indian Ocean region, in order to understand

their origin and history.

Location Western Indian Ocean region.

Methods We analysed sequences of mitochondrial DNA (partial 12s and 16s rRNA genes, 766 bp) from 48 specimens collected in Madagascar, Mauritius, the four Comoros islands and East Africa, and also in New Caledonia, representing the Australo-Pacific unit of the distribution.

Results Pairwise sequence divergences of c. 3.1% were found between the New Caledonian forms and the ones from the Western Indian Ocean. Two clades were identified in Madagascar, probably corresponding to the recognized forms

cognatus and voeltzkowi, and two clades were identified in the Comoro islands,

where each island population formed a distinct haplotype clade. The East African samples form a monophyletic unit, with some variation existing between Pemba, Zanzibar and continental Tanzania populations. Individuals from Mauritius form a divergent group, more related to populations from Moheli and Grand Comore (Comoros islands) than to the others.

Main conclusions The level of divergence between the populations from the Western Indian Ocean and Australian regions and the geographic coherence of the variation within the Western Indian Ocean group are concordant with the hypothesis of a colonization of this region by a natural transoceanic dispersal (from Australia or Indonesia). The group then may have diversified in Madagascar, from where it separately colonized the East African coast, the Comoros islands (twice), and Mauritius. The genetic divergence found is congruent with the known morphological variation, but its degree is much lower than typically seen between distinct species of reptiles.

Keywords

Africa, Comoros, Cryptoblepharus, island colonization, Madagascar, Mauritius, Scincidae, Squamata, transoceanic dispersal.

S. Rocha ef a/.

I N T R O D U C T I O N

Cryptoblepharus is a panPacific genus of c. 36 morphologically

similar skinks, formerly considered as subspecies of a single variable species, Cryptoblepharus boutonii (Mertens, 1934; Greer, 1974). These lizards occur in two disjunct areas: (1) the eastern end of the IndoAustralian archipelago, Australia and Oceania; and (2) islands of the far Western Indian Ocean and adjacent parts of the African coast (Fig. la, adapted from Greer, 1974 and Branch, 1988).

Some attempts have been made to unveil the origin and biogeographical history of Cryptoblepharus. Mertens (1931) suggested that an ancestral form evolved in Southeast Asia and migrated to Australia, where the genus evolved and diversified, and, by passive means of dispersal, radiated to its present distribution in the Australian and IndoPacific regions. Greer (1974), analysing the intergeneric relationships of Cryptoblepharus, suggested that the most primitive

Cryptoblepharus could be derived from Emoia, another genus

with a wide distribution in the Pacific region and another obvious excellent transmarine disperser. Neither of these authors, however, offered any explanatory hypothesis for the disjunct Western Indian Ocean distribution. Biogeographical theory indicates three mechanisms by which Cryptoblepharus could have achieved its disjunct, widespread distribution: the evolution of an ancestral form at a time when the regions were connected; natural longdistance transoceanic dispersal over a long period of time; or recent humanmediated transportation. Judging from other studies of skinks (Carranza & Arnold, 2003), Cryptoblepharus is not sufficiently ancient for a Gondwanan origin, and the recognized morphological variation within the genus supports the pre human occurrence of Cryptoblepharus across its distribution range. It is therefore likely that Cryptoblepharus has had a long period of evolution and has naturally dispersed to large and small islands, rafting on driftwood or vegetation mats, with some islands possibly being used as 'stepping stones' to colonize others, a pattern already identified in archipelagos

such as Hawaii (Gillespie, 2002, 2004) and the Canary islands (Thorpe et al, 1994; Brown & Pestano, 1998). In fact, many natural attributes of Cryptoblepharus, such as low metabolic requirements, ectothermy, frequent occurrence on marine beaches and adaptation to habitats devoid of fresh water could facilitate the crossing of open seas (e.g. Fricke, 1970). These oceanic dispersals could have been as extensive as the c. 6000 km that separates the IndoAustralian and the Western Indian Ocean regions, and may explain the colonization of Madagascar, East Africa and the surrounding islands by

Cryptoblepharus. In fact, recent evidence has led to a

resurrection of the dispersal hypothesis in historical bio geography, indicating that such dispersal events may have been more common than previously thought by vicariance biogeographers (de Queiroz, 2005).

In the last extensive review of this genus in the Western Indian Ocean, Brygoo (1986) recognized 13 forms, with allopatric distributions (Fig. lb). With the exception of eleva ting the Europa island form, C. b. bitaeniata, to a specific status, and proposing the treatment of C. b. mayottensis (from Mayotte, Comoros) and C b. mohelkus (from Moheli, Comoros) as varieties of the subspecies C. b. gloriosus (from Glorioso island), he maintained the subspecific treatment given to the forms by previous authors. P. Horner (unpublished data), in a multivariate analysis of morphological data, identified 13

Cryptoblepharus taxa from the Western Indian Ocean region, 12

of which were distinguished by two or more statistically significant morphological differences. He considered all 13 taxa to be separate 'species' (P. Horner, pers. comm.).

The complexity in assigning a taxonomic status to these forms comes from their longappreciated distinctive colour patterns together with their distribution: there is a good deal of variation between the forms (Mertens, 1931), but since most of them occur on separate islands, their true status is beyond the test of sympatry (Greer, 1974). Thus it is unclear if there exists one polytypic species or several distinct species, a problem that can now be evaluated using a molecular approach. •

₃₄₁₁

_"_V_{""W< / ~ ^ s y}

^<>ays

? t

w

y '

/ * r ^

ÒB

*>%

*_{\ ***r}*

^ ^

\5/

V H ^J

f *

\J2%

Grand Comoro Gtorioeo c. à *tor ■ Anjouan ■c- *■ * > * « Moheli \ * a C U k * « A Ch.mohtkM * _<JNÒXyBe Mayotto C 6. nê/oHtn Mozambfcue h .

Juan de Nova Is. C. hcMdMuc

C. b. MMITIWB

Mauritius

Figure 1 (a) Distribution of the genus Cryptoblepharus (adapted from Greer, 1974 and Branch, 1988). (b) Distribution of the Western Indian Ocean subspecies (from Brygoo, 1986).

Here we examine Cryptoblepharus from the Western Indian Ocean region, including eight recognized forms from Madagascar (two), the Comoros archipelago (four), East Africa (one), and Mauritius Island (one). Cryptoblepharus

novocaledonicus, from New Caledonia, was also included

in the analysis. On the basis of mitochondrial DNA sequences we elucidate the phylogenetic relationships among these forms and address possible historical vicariant and dispersal patterns that may have shaped their current distribution.

MATERIALS AND METHODS

Tissue samples (tail tips) were collected in various localities (see Table 1 and Fig. 2) across Madagascar, Mauritius, the four major islands of the Comoros archipelago, the East African coast (mainland Tanzania, Zanzibar and Pemba islands) and New Caledonia and preserved in 98% ethanol. Total genomic DNA was extracted using standard high-salt protocols (Sam- brook et al, 1989), and fragments of the 12s and 16s rRNA genes were amplified using universal primers (12Sa and 12Sb from Kocher et al, 1989 and 16Sar-L and 16Sbr-H from Palumbi et al., 1991) and following Harris et al. (1998). The PCR products were sequenced in an automated DNA sequencer (ABI PRISM 310) following the manufacturer's instructions. GenBank accession numbers for new sequences are DQ118039-DQ118080. For one individual from Madagas- car, sequences of both genes were already available on GenBank (Schmitz et ai, 2005). Sequences were aligned manually using BIOEDIT (Hall, 1999), and genes were com- bined, resulting in a 766-bp fragment. Within the Cryptoble-

pharus group the alignment was unambiguous as only single

indels were included. The Lygosominae skinks Leiolopisma

telfairi (Carranza & Arnold, 2003) and Emoia cyanura

(Whiting et al, 2003) were included as outgroups. A short hypervariable region (15 bp) was removed for the analysis with these outgroups.

Maximum likelihood (ML), maximum parsimony (MP) and Bayesian analyses were performed. For the ML and Bayesian analysis, the model of nucleotide substitution that best fits our data set was selected using Modeltest 3.06 PPC (Posada & Crandall, 1998), under the Akaike information criterion (following Posada & Buckley, 2004). For the combined data set, the best-fitting model was the General Time Reversible (GTR + I + G), with base frequencies and substitution rates estimated from the data, a proportion of invariable sites of 0.6713, and a gamma distribution shape parameter of 0.7343. The software PHYML (available at http://www.lirmm.fr/w3ifa/ MAAS/), which implements an algorithm that adjusts tree topology and branch lengths simultaneously (Guindon & Gascuel, 2003), departing from an initial tree constructed using BIONJ (Gascuel, 1997), was used to perform the ML

geography of Cryptoblepharus in the Western Indian Ocean data partition by gene, applying an appropriate model for each gene (also selected using Modeltest and the Akaike information criterion). These were the TVM + G for 12s and GTR + I for 16s.

The application of these models resulted in a distorted tree topology, however, probably as a result of the high distance of the outgroups compared with all Cryptoblepharus. The esti- mated proportion of invariable sites was very low (67% for the combined fragment, and 80% for the 16s only), causing the estimate of topology to be incorrectly rooted (see results). Thus, we applied the model to both combined and partitioned-by-gene Bayesian analysis, excluding the proportion of invariable sites (/) parameter. Parameters were estimated as part of the analysis with four Markov chains incrementally heated with the default heating values. All analyses started with randomly generated trees and ran for 2 x 10s generations,

saving one tree in every 100 generations. The log-likelihood values of the sample points were plotted against the generation time and all the trees prior to reaching stationarity were discarded, making sure that burn-in samples were not retained. Combining the remaining trees, a 50% majority rule consensus tree was generated. The frequency of any particular clade of the consensus tree represents the posterior probability of that clade (Huelsenbeck & Ronquist, 2001). Maximum parsimony (MP) analysis was also carried out, in PAUP* 4.0bl0 (Swofford, 2002), using heuristic searches involving tree bisection and reconnection (TBR) branch swapping, with 100 replicates. Gaps were considered as a fifth state and all characters were weighted equally. Robustness of these trees was assessed by bootstrap analysis (Felsenstein, 1985) involving 1000 pseudo- replications.

Because the level of divergence within the Cryptoblepharus sequences was low, a median-joining network (Bandelt et al, 1999) was constructed using NETWORK software (Fluxus Engineering, Suffolk, UK) for these sequences only. Networks of interconnected haplotypes represent the evolutionary relationships and gene genealogies within species better than the bifurcating patterns usually recovered by methods of phylogenetic inference (Posada 8c Crandall, 2001).

RESULTS

We obtained a total of 48 Cryptoblepharus sequences for both the 12s and 16s gene fragments. Maximum parsimony analysis recovered 16 equally most parsimonious trees (180 steps; consistency index 0.872; retention index 0.910). The strict consensus tree of the MP tree topologies was identical to the ML tree, although less resolved. In the Bayesian analysis of data, both considering one model for the combined fragments (GTR + G + I) and one independent model for each gene (TVM + G for 12s and GTR + I for 16s), the tree topology was clearly distorted, with the outgroup rooting the tree with

S. Rocha et al.

Table 1 List of samples used for analysis, geographic locations, and GenBank accession numbers for 12s and 16s

Species Specimen Locality Island; country Accession numbers

C. b. cognatus 2000/MB11 Ambolobozaleki, south Diego Madagascar DQ118039/DQ118060

C. b. cognatus 2000/MB56 Nosy Fanihy Madagascar DQ118040/DQ118061

C. b. cognatus 2000/MB32 Nosy Fanihy Madagascar DQ118040/DQU8061

C. b. cognatus 2000/MC5 Nosy Fanihy Madagascar DQ118040/DQ118061

C. b. cognatus 2000/MC9 Nosy Sakatia Madagascar DQ118041/DQ118062

C. b. cognatus 2000/MB61 Nosy Sakatia Madagascar DQ118041/DQ118062

C. b. cognatus 2000/MB43 Nosy Be Madagascar DQ118042/DQ118063

C. b. voeltzkowi 2000/M520 Ifaty, SW Madagascar Madagascar DQ118052/DQ118073

C. b. voeltzkowi FGZC78 Lebanona Beach, Tolagnaro Madagascar DQU8053/DQ118074

C. b. voeltzkowi FGZC79 Lebanona Beach, Tolagnaro Madagascar DQ118053/DQU8074

C. b. voeltzkowi FGZC80 Lebanona Beach, Tolagnaro Madagascar DQ118053/DQ118074

C. b. voeltzkowi FGZC84 Lebanona Beach, Tolagnaro Madagascar DQ118053/DQU8074

C. b. voeltzkowi FGZC87 Lebanona Beach, Tolagnaro Madagascar DQ118053/DQ118074

C. b. voeltzkowi CbvE25 St. Augustin, near Ianantsony Madagascar AY308336/AY308219

C. b. mayottensis MY68 Mamoutzou, harbour Mayotte; Comoros DQ118043/DQ118064

C. b. mayottensis MY69 Mamoutzou, harbour Mayotte; Comoros DQ118042/DQ118063

C. b. mayottensis MY70 Mamoutzou, harbour Mayotte; Comoros DQ118044/DQ118065

C. b. degrisii AJ4 Moutsamoudu Anjouan; Comoros DQ118047/DQ118068

C. b. degrisii AJ6 Moutsamoudu Anjouan; Comoros DQ118046/DQ118067

C. b. degrisii AJ20 Moya Anjouan; Comoros DQ118045/DQ118066

C. b. degrisii AJ21 Moya Anjouan; Comoros DQ118045/DQ118066

C. b. mohelicus MH16 Fomboni, harbour Moheli; Comoros DQ118054/DQ118075

C. b. mohelicus MH17 Fomboni, harbour Moheli; Comoros DQ118054/DQ118075

C. b. mohelicus MH18 Djayézi Moheli; Comoros DQU8054/DQ118075

C. b. ater GC59 Moroni, harbour Grand Comore; Comoros DQ118056/DQ118077

C. b. ater GC60 Moroni, harbour Grand Comore; Comoros DQ118056/DQ118077

C. b. ater GC61 Moroni, harbour Grand Comore; Comoros DQ118056/DQ118077

C. b. ater GCh Gouni, Sandini Grand Comore; Comoros DQ118056/DQ118077

C. b. ater GCi Gouni, Sandini Grand Comore; Comoros DQ118056/DQ118077

C. b. ater F893 Chomoni beach Grand Comore; Comoros DQU8055/DQ118076

C. b. boutonii Maur Gabriel island Mauritius, Mascarenes AF280116/AY151445

C. b. boutonii Ma623 lllc de la Passe Mauritius, Mascarenes DQ118057/DQ118078

C. b. boutonii Ma624 Ille de la Passe Mauritius, Mascarenes DQ118057/DQ118078

C. b. africanus Z6 Stone town, W Zanzibar, Tanzania DQ118050/DQ118071

C. b. africanus Z7 Stone town W Zanzibar, Tanzania DQ118050/DQ118071

C. b. africanus Z8 Stone town W Zanzibar, Tanzania DQ118050/DQ118071

C. b. africanus Z9 Stone town W Zanzibar, Tanzania DQ118050/DQ118071

C. b. africanus Z32 Kiwengwa E Zanzibar, Tanzania DQU8051/DQ118072

C. b. africanus Z45 Mena Bay SW Zanzibar, Tanzania DQ118051/DQ118072 C. b. africanus Z46 Mena Bay SW Zanzibar, Tanzania DQ118051/DQ118072

C. b. africanus PB13 East Chake Pemba, Tanzania DQ118048/DQ118069

C. b. africanus PB14 East Chake Pemba, Tanzania DQ118048/DQ118069

C. b. africanus PB16 Jondeni SW Pemba, Tanzania DQ118049/DQ118070

C. b. africanus PB28 Ngezi N Pemba, Tanzania DQ118049/DQ118070

C. b. africanus TZ46 DarEsSallam, Msanani penins. continental Tanzania DQ118050/DQ118071

C. b. africanus TZ47 DarEsSallam, Msanani penins. continental Tanzania DQ118050/DQ118071 C. novocaledonicus AMB7210 Isle of Pines New Caledonia DQ118059/DQ118080

C. novocaledonicus AMB8050 Loyalty island New Caledonia DQU8058/DQ118079

Leiolopisma telfairi Round Island Mauritius AF280122/AY151450

Emoia cyanura Vitilevu, Sigacota Fiji AY218018/AY217968

concordant with the previous ones. The ML tree, with ML and MP bootstrap values and Bayesian posterior probabilities of data partitioned analysis, is represented in Fig. 3.

Cryptoblepharus from the Western Indian Ocean region

form a monophyletic unit with c. 3.1% pairwise divergence

(average between group uncorrected P-distance) in relation to C. novocaledonicus (more than 20 mutational steps). Among taxa from the Western Indian Ocean region, pairwise sequence divergences ranged from 0% to 2.4%. A detailed analysis of the distribution of the haplotypes within 16 Journal of Biogeography 33, 13-22, © 2005 Blackwell Publishing Ltd

Biogeography of Cryptoblepharus in the Western Indian Ocean

Maur623 MaurBM .

Ù Mauritius

Figure 2 Sampling localities in Tanzania (east Africa), the Comoros archipelago, Madagascar and Mauritius (grey dots), and haplotype median-joining network of the analysed subspecies of Cryptoblepharus boutonii, using combined 12s and 16s rRNA gene fragments. Each circle corresponds to one observed haplotype, with the size of the circle proportional to the number of individuals in which the haplotype was found (inside). Small black dots represent missing haplotypes.

Cryptoblepharus (Fig. 2) revealed clear geographical structur-

ing. Madagascar harbours the highest haplotype diversity, as seven haplotypes were found from seven localities, with as many as 14 mutations between them. These can be placed into two putative groups distributed in north-western Madagascar, (1) Nosy Be and surrounding islets and (2) southern Madagascar. The haplotypes from the individuals

other formed by the populations from Moheli and Grand Comore. The haplotypes from Mayotte and Anjouan are clearly derived from the north-western Madagascar ones, while the origin of the Moheli and Grand Comore ones, with just one mutational step between them, cannot be ascertained with certainty.

S. Rocha e t al. «a nm 12« Madagascar Africa

,¾

PB1S peu peia Muau MV70 una

4L-

a . Madagascar Comoros Madagascar Comoros Od 00(11 UHtT una —iòniiMi gniOOl AUB72I0 Comoros Mauritius New Caledonia

Figure 3 Maximum likelihood phylogram of all the Cryptoblepharus individuals sequenced for 12s and 16s rRNA. Emoia cyanura and

Leiolopisma telfari are used as outgroups. Individual codes are as used in Table 1 and Fig. 2. Bootstrap values for MP and ML and Bayesian

posterior probabilities (PP) above 50% are shown (in percentage: ML, MP/PP). ongoing or past hybridization between the two forms is a

matter for future investigation.

Despite the few individuals analysed, more haplotypes were detected within Anjouan and Mayotte (three in each island) than in Moheli and Grand Comore (one and two, respectively). This is congruent with the older age of the first two islands. In fact, these hotspot-originated volcanic islands span a wide range of ages: Mayotte, 10-15 Myr;

Anjouan, 11.5 Myr; Moheli, c. 5 Myr; and Grand Comore 0.5 Myr - age of the oldest exposed lavas in the case of Moheli and estimated age for the volcanic origin of the other three Islands (Montaggioni & Nougier, 1981; Emerick & Duncan, 1982; Nougier et ai, 1986). The haplotypes from Mauritius are closest to those from Moheli and Grand Comore, but have a relevant genetic differentiation of 10 mutational steps.

Biogeography of Cryptoblepharus in the Western Indian Ocean Concerning the east African populations, some variation

was observed: two closely related haplotypes were detected in individuals from Zanzibar and mainland Tanzania (with sharing of haplotypes between west of Zanzibar and mainland Tanzania), and two other slightly divergent ones were detected in Pemba populations (the other small offshore island). All these east African haplotypes form a monophyletic unit derived from the southern Malagasy haplotypes.

DISCUSSION

Origin of Cryptoblepharus in the Western Indian Ocean

As stated in the Introduction, three mechanisms can be invoked to explain the origin of Cryptoblepharus in the Western Indian Ocean region: (1) ancient vicariance, (2) transoceanic dispersal, and (3) translocation by humans. Our data provide novel information to assess these hypotheses. If vicariance was the origin of the disjoint Cryptoblepharus distribution, the separation of the main Cryptoblepharus populations would be related to the break-up of Gondwana in Mesozoic times. Much higher genetic divergences than the 3.1% pairwise divergences would be expected between

C. novocaledonicus and the Western Indian Ocean forms

under such a scenario. If human-mediated transportation were responsible for the presence of Cryptoblepharus in the Western Indian Ocean region, no geographically structured genetic variation at these mitochondrial markers would be expected, as in the case of some Hemidactylus species from this same region (Vences et al, 2004b; Rocha et al, 2005), or in introduced Polynesian lizards (Austin, 1999). Hence, our data strongly support an origin of Cryptoblepharus in the Western Indian Ocean region by natural transoceanic dispersal. Because no close relatives of Cryptoblepharus occur in the Western Indian Ocean, we assume that the direction of dispersal was from the Indo-Pacific region towards Madagascar.

Despite the lack of Australian and Indonesian taxa in this study, given the relatively low degree of variation among populations from the Western Indian Ocean and the divergence observed between these and the New Caledonian individuals, the most parsimonious hypothesis is to assume only one colonization event to the Western Indian Ocean, probably to Madagascar with subsequent dispersal to surrounding islands. Furthermore, the hypothesis of two colonization events, by individuals belonging to very closely related lineages, is much less likely.

The data therefore suggest a geologically recent longdistance overwater dispersal followed by several minor dispersal events. Variation between the two major Malagasy haplotype clades is at least 1.82%. Assuming a rRNA evolution rate of 0.625% per Myr since the last common ancestor (Lin

A similar pattern is suggested for several other taxa, such as the gekkonids Nactus and Lepidodactylus and the Leiolopisma skinks (Austin et al, 2004), which arose in Southeast Asia and reached the Mascarenes archipelago via the west-running Equatorial current. Several other long-distance transoceanic dispersals are known in reptiles: the ancestor of Phelsuma

No documento Phylogeny, Phylogeography and colonization patterns of reptiles in Indian Ocean Islands (páginas 42-70)

3411

^<>ays

w

y ­ '

ÒB

\ ***r

\5/

V H ^J

f *

\J2%

,¾

4L-

₃₄₁₁

y '

*_{\ ***r}*